From LWN: “Linus has just merged the XFS filesystem into his BitKeeper tree; it will thus show up in the 2.5.36 kernel. XFS is a high-performance, journaling filesystem from SGI; it now becomes the fourth journaling filesystem (alongside ext3, ReiserFS, and JFS) supported by the Linux kernel. (Other stuff which has been merged, so far, for 2.5.36 includes an IEEE-1394 (“Firewire”) update, the next big set of IDE patches, the “huge page” patch for i386 systems, and a number of other tweaks).”
Linus Merges XFS on Kernel 2.5.36
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Hip-hip Hooooray!!!! 🙂
-fooks
what does ‘huge page’ mean? compare to ‘super pager’ http://shimizu-lab.dt.u-tokai.ac.jp/lsp.html
greetings, MaHejn
finaly
I wonder when 2.6 will be out… *drooling*
FBSD 5 vs Linux 2.6 will be an interesting match
The dev kernel went up to number 2.3.51
But when 2.6 is out nobody really knows yet.
Guesses by kernelhackers have been 2.6 out next summer.
When it is ready. 🙂
I can’t wait for the feature freeze, that way we can roughly guess when 2.6/3.0 is coming out.
Anyway, I just tried XFS two days ago, boy was it fast. Still too early to jugde its stablity, but if it is as stable as ext3, XFS would be the FS of my heart :-). (Besides, thanks to the experiment, I ruin my partition table, albeit manage to save my MP3/Ogg collection… it would be weeks till I feel like fixing things up)
>Anyway, I just tried XFS two days ago, boy was it fast
I didnt notice and noteable dffrences compared to ext3.
The biggest thing about XFS is imho that it supports ACL/EAs
meaning you can give arbitarey access rights to as many users and groups you want on a file.
which is, in your well-informed opinion, the best journalling FS? I know this will probably sound like a flame-bait, but believe me, I have to know.
My job (as well as other’s people) literally depends on my hardisk (yes, i do backups, but even with my daily backup I risk missing quite a lot of data), and i want to be sure that I can recover as much data as possible if (god forbids) something happens to my harddisk.
I am currently running SuSE 8.0 with reiserfs and i was pretty upset by comments on the suse mailing list about people having truble restoring their reiserFs FS after a crash ….
i would really appreciate your opinion
I think for safety you’re best off with ext3 with most rigorous settings (data=ordered). However, don’t expect to get any decent performance out of it.
I read a few things on /. about XFS where it corrupts more easily than reiserfs and ext3 if the power is pulled on a running system.
For me personally, I think XFS seems a bit bloated for my taste. I’d also say that reiserfs seems to be the most cutting edge of the linux file systems, it will probably benefit most from the upcoming file systems wars.
Will openbfs be a viable option for a native linux file system? Or will someone port over a bsd style soft updates type file system and add that to the mix?
http://www-106.ibm.com/developerworks/linux/library/l-fs7/
And yeah its the simplest to set up (providing your on a Linux box). Just tunefs -j /dev/<your hd> and the appropriate modifications to your fstab.
Some of the annoying things I’ve noticed about ext3 tho is that its not supported by OpenBSD (where as XFS is) and if you accidentally delete data you can kiss it goodbye (the traditional ext2 recovery tools don’t work because of the way its set up). Don’t know about data recovery on XFS.
Definately XFS, except for certain cases. XFS is generally faster than ReiserFS for medium to large sized files, extremely stable (has a better track record than ReiserFS overall) and has a lot of very useful features (not only ACLs and attributes, but you can resize mounted filesystems, do live filesystem dumps, and the recovery tools are great). As for technical superiority, I’d have to give the edge to XFS here as well. ReiserFS is different (the whole balanced tree model vs the traditional i-node model) but XFS still has some features (delayed block allocation, B+ tree based allocator, realtime volumes, etc) that Reiser doesn’t. Reiser-3 also has some implementation problems (like only being able to write 4KB for each traversal of the tree) that won’t get fixed until version 4. The only place I see Reiser being better, at the moment anyway, is in situations with small files. I’m setting up a version control server here, and since it will be dealing with lots of small files (source code diffs) it’ll be running ReiserFS. That said, I’m looking forward to Reiser4. Hans has a bunch of tricks up his sleeve for the next iteration.
Instant deletion. If you do alt.binary newsgroup downloads, you accumulate a lot of multipart archives and huge files. I like being able to type “rm Thingy*” and immediately have my prompt back, without massive disk activity.
Just the fact that the filesystem doesn’t have to go through and write to every single inode (at every 1kb, 2kb, or 4kb) every time a change is made will greatly decrease the chance for fault. In ext2, those inodes always remain in those spots, and are constantly being overwritten whenever a change is made. That’s an excellent recipe for creating bad blocks.
I haven’t tried ReiserFS, due in large part to the “help” info in make config. “ReiserFS is made by elitedudes, We are elite, we make ReiserFS. If you like ReiserFS, you can pay us monies. If you want features, you pay use moneys and we’ll add it 4 u!!! We are elite dudes. Visit us on the interweb at http://www.eliteness.com!!!!!!!!” Mind you, that’s vaguely a paraphrase. That’s the impression I get from it, and I sincerely dislike it. Apparently there are also massive compatibility problems with ReiserFS’ infinite versions. I’ve heard nothing but bitching about incompatibly changing the disk format every few weeks, when someone wants a new feature. That stinks of bad design, feature creep, and unreliable developers. Some would call those political reasons, I call it good sense.
anon: I didnt notice and noteable dffrences compared to ext3.
If you transfer a lot of files, I think you can see a noticible increase of performance. Like transfering almost 1,000 Ogg/MP3 files.
fred: which is, in your well-informed opinion, the best journalling FS?
The fastest I know is XFS. It is also I heard one of the most unstable.
But then you are better off with ext3.